An Anthropic researcher quit saying AI labs are gambling with our lives

Summary

Anthropic pretraining researcher Jacob Coxon resigned over concerns that leading laboratories prioritize competitive races over safety. Evan Hubinger, Anthropic's head of alignment science, publicly acknowledged these alignment challenges and cited personal estimates of existential risk.

Key Takeaways

1. Jacob Coxon, a 27-year-old pretraining researcher who had spent three years at OpenAI and then Anthropic, announced his resignation on X on Tuesday.
2. Evan Hubinger, who leads alignment science at Anthropic, posted on X the following day: “Jacob is correct here; we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade.”
3. “I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to.”

Statement Breakdown

• Claimed Facts: 45% of statements the article presents as facts
• Opinions: 35% of statements classified as editorial or subjective
• Claims: 20% of statements surfaced for additional reader evaluation

Credibility & Bias Reasoning

Credibility assessment: The reporting relies on public statements posted on social media by identifiable corporate insiders. It provides context on internal industry dynamics while referencing previous regulatory actions and executive statements. Several broader claims concerning competitive dynamics and catastrophic risk levels reflect individual perspectives rather than verified industry consensus.

Bias assessment: Tech Industry Safety Skepticism. The commentary exhibits a critical posture toward private frontier artificial intelligence laboratories. It portrays corporate safety commitments as secondary to commercial competitive pressures. The analysis validates resignations and warnings from safety personnel while framing commercial leadership as trapped in competitive escalation.

Note: Contains direct commentary and subjective probability estimates from industry insiders alongside factual reporting.

Credibility flag: Direct Insider Statements

Claimed Facts (5)

• This statement describes a factual career move and public announcement made on a specific platform.
• This states the specific factual nature of Coxon's departure from the artificial intelligence sector.
• This records the verifiable timing and role of a named employee posting on social media.
• This catalogs a chronological sequence of corporate policy shifts and governmental interventions.
• This reports the factual absence of a formal communication from the company.

Opinions (5)

• This expresses a subjective evaluation of corporate responsibility.
• This conveys subjective judgments regarding safety standards and private governance.
• This reflects a personal assessment of organizational readiness and strategic progress.
• This provides editorial framing regarding the competitive dynamics of the industry.
• This summarizes an inferred collective mindset without empirical measurement.

Claims (5)

• This makes an unprovable, emotionally charged claim about existential jeopardy.
• This presents a subjective probability of human extinction that cannot be empirically validated.
• This amplifies an unquantifiable existential estimate as an authoritative corporate risk metric.
• This asserts widespread technical dissent using non-specific references.
• This uses rhetorical assertion to justify treating subjective probability statements as major factual developments.

Key Sources

• Ana-Maria Stanciuc — Author at The Next Web
• Jacob Coxon — Former Pretraining Researcher at Anthropic and OpenAI
• Evan Hubinger — Alignment Science Lead at Anthropic